Model for Cyclic Inelastic Buckling of Steel Braces

Patxi Uriz1; Filip C. Filippou2; and Stephen A. Mahin3

Abstract: The paper presents a model for the inelastic buckling behavior of steel braces. The model consists of a force-based frame
element with distributed inelasticity and fiber discretization of the cross section. With this approach, the response of the element can be
derived by integration of the uniaxial stress-strain relation of the fibers and can account for kinematic and isotropic hardening as well as
the Bauschinger effect of the material. The interaction between axial force and bending moment is thus accounted for. Even though the
Downloaded from ascelibrary.org by Kookmin University on 10/14/22. Copyright ASCE. For personal use only; all rights reserved.

element only accounts for small deformations in the basic system, large displacement geometry is included in the nonlinear transformation
of the force-deformation relation of the basic element following the concept of the corotational formulation. With this approach, two
elements for each brace suffice to yield results that match an extensive set of experimental data of braces with different cross sections and
slenderness ratios. Even though the model does not account for the effect of local buckling, this phenomenon does not seem to appreciably
affect the global response of braces with compact sections.
DOI: 10.1061/共ASCE兲0733-9445共2008兲134:4共619兲
CE Database subject headings: Inelasticity; Buckling; Bracing; Steel.

Introduction ond order bending moment and the axial force; and 共c兲 three-
dimensional finite element models.
Concentrically braced steel frames are commonly used in new Phenomenological models are the simplest and most computa-
earthquake resistant structures or for the seismic retrofit of exist- tionally efficient; in these, the brace is represented by a truss
ing structures. The seismic performance of this structural system element with hysteretic behavior that mimics the experimentally
is sensitive to the hysteretic behavior of the braces 共Uriz and observed response. The limitation of this approach is the need for
Mahin 2004兲. The lateral response of concentrically braced steel model calibration against available experimental data, and the
frames is characterized by relatively large stiffness, so that inter- limited predictive ability, since the hysteretic behavior only rep-
story drift requirements are easily satisfied under service loads. resents the behavior of specimens for which it was calibrated.
As a result, there is little overstrength in the lateral capacity of the Nevertheless, phenomenological models have been used with suc-
structural system compared to its steel moment frame counterpart, cess in several studies 共Ikeda et al. 1984; Khatib et al. 1988;
and large inelastic excursions should be expected under the most Zayas et al. 1980兲.
severe ground motion at the site. At the opposite end of the spectrum are three-dimensional fi-
The hysteretic response of steel braces has been the subject of nite element models that derive the hysteretic response of the
numerous experimental studies 共Black et al. 1980; Gugerli and brace from the nonlinear material response under large deforma-
Goel 1980; Lee and Goel 1987; Liu and Goel 1988兲. In these tion theory. Such models are not commonly used in structural
studies, various types of cross section have been tested including engineering applications, because of their complexity and compu-
concrete-filled tubes. These studies addressed the effect of slen- tational expense.
derness on overall buckling, the effect of width-to-thickness ratio Significant effort to date has been expended in the second
on local buckling, as well as the effects of end restraint and low category of beam-column modeling for which 共Ikeda and Mahin
cycle fatigue on brace behavior. 1986兲 have used the term “physical theory model.” Most of these
Previous analytical models for steel braces fall into three cat- models are based on a linear elastic beam-column element with
egories: 共a兲 phenomenological; 共b兲 beam-column elements with an inelastic hinge at midlength of the brace. Thus, end restraint
various approaches for describing the interaction between the sec- conditions cannot be represented. Hall and Challa 共1995兲 used a
fiber discretization of the cross section and some form of the
1 corotational formulation to describe the inelastic hysteretic re-
Exponent, Failure Analysis Associates, Menlo Park, CA 94025.
E-mail: puriz@exponent.com sponse of steel members, but the correlations with experimental
2
Civil and Environmental Engineering, Univ. of California, Berkeley, results are limited to a single specimen and the quality of the
CA 94720. E-mail: filippou@ce.berkeley.edu correlation was not very good. More recently, Jin and El-Tawil
3
Civil and Environmental Engineering, Univ. of California, Berkeley, 共2003兲 have used a distributed inelasticity element with a bound-
CA 94720. E-mail: mahin@ce.berkeley.edu ing plasticity model of force resultants for the interaction between
Note. Associate Editor: Donald W. White. Discussion open until bending moment and axial force. A geometric stiffness matrix was
September 1, 2008. Separate discussions must be submitted for individual added to the material stiffness matrix, but no further details were
papers. To extend the closing date by one month, a written request must
provided. The correlations of the model with experimental evi-
be filed with the ASCE Managing Editor. The manuscript for this paper
was submitted for review and possible publication on May 31, 2005; dence show reasonable accuracy in the tensile stress range, but
approved on October 25, 2005. This paper is part of the Journal of are less accurate under compression forces.
Structural Engineering, Vol. 134, No. 4, April 1, 2008. ©ASCE, ISSN The purpose of this paper is to demonstrate the capability of an
0733-9445/2008/4-619–628/$25.00. inelastic beam-column element model within the OpenSees com-

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2008 / 619

J. Struct. Eng., 2008, 134(4): 619-628

Downloaded from ascelibrary.org by Kookmin University on 10/14/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Transformation of frame element forces and deformations from basic to global reference system under nonlinear geometry of large
displacements

putational framework of the Pacific Earthquake Engineering Re- body modes derives from small deformation theory. Nonlinear
search Center 共McKenna 1997兲 共http://opensees.berkeley.edu兲 in geometry under large displacements can be accounted for during
modeling the hysteretic response of steel braces. The inelastic the transformation of the element forces and deformations from
beam-column element of this study is limited to small deforma- the basic to the global reference system of the structural model, as
tions in the basic system and is based on the formulation by will be briefly described in the following for the 2D element
Spacone et al. 共1996兲. The element accounts for large displace-
formulation. Details about 3D geometric transformation can be
ments by embedding the basic system in a corotational frame-
found in de Souza 共2000兲.
work, as proposed by R. de Souza in his Ph.D. thesis 共de Souza
2000兲. de Souza also developed a new inelastic frame element Fig. 1 shows the inelastic frame element in its original and
with moderate deformations in the basic system, which offers the deformed configuration. The element forces in the basic system
advantage of a coarser element discretization for capturing local are denoted with q and the corresponding deformations with v.
deformations in buckling members. This element will be dis- The element forces and displacements in the local reference sys-
cussed in a forthcoming paper. tem x − y are denoted with p̄ and ū, respectively. Finally, the
The validity of the proposed modeling approach for steel forces and displacements in the global reference system X − Y of
braces is established through an extensive set of correlation stud- the structural model are denoted with p and u, respectively. For-
ies with steel braces of different cross-section shapes. The culmi- mulating the equilibrium in the basic system that is assumed to
nation of this extensive study is a set of modeling guidelines for follow the deformed element configuration according to the coro-
the representation of this type of structural member in
tational theory gives the following relation between basic forces q
performance-based earthquake engineering applications.
and end forces p̄ in the local reference system 共Filippou and
Fenves 2004兲:

冢冣冤 冥
Analytical Model
共L + ⌬ūx兲 ⌬ūy ⌬ūy
The inelastic frame element in this study accounts for distributed − − −
Ln L2n L2n
inelasticity through integration of material response over the
cross section and subsequent integration of section response along
p̄1 ⌬ūy L + ⌬ūx L + ⌬ūx
−
the element. It is based on the force formulation 共Spacone et al. p̄2 Ln L2n L2n
1996兲, an approach that offers significant advantages over the p̄3 0 1 0
more common displacement formulation: 共a兲 the force- p̄ = =
p̄4 L + ⌬ūx ⌬ūy ⌬ūy
interpolation functions are always exact in the absence of second
order effects; 共b兲 a single element can be used to represent the p̄5 Ln L2n L2n
curvature distribution along the entire member with sufficient ac- p̄6 ⌬ūy L + ⌬ūx L + ⌬ūx
curacy through selection of a sufficient number of integration − −
Ln L2n L2n
points 共monitoring sections兲; and 共c兲 the formulation has proven
numerically robust and reliable, even in the presence of strength 0 0 1
softening, as is the case for buckling steel braces. This element is

冢冣
in extensive use in seismic response simulation studies within the q1
OpenSees framework.
⫻ q 2 = b uq 共1兲
The frame element response in the basic system without rigid
q3
620 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2008

J. Struct. Eng., 2008, 134(4): 619-628

 where Ln⫽deformed element length; and ⌬ū⫽relative displace-
ment in the local reference system 共Fig. 1兲. The end forces can be
transformed to the global reference system with a rotation trans-
formation matrix ar in the usual manner. We can, thus, write
p = arTp̄ 共2兲
The tangent stiffness matrix of the element at its current state
can be obtained by differentiation of the end forces p with respect
to the corresponding end displacements. The first step of the pro-
cess with the aid of Eq. 共2兲 is trivial
⳵p ⳵ p̄ ⳵ ū ⳵ p̄
ke = = arT = arT ar 共3兲
⳵u ⳵ ū ⳵ u ⳵ ū
Fig. 2. Experimental hysteretic response of 4 in. extra-strong pipe
Downloaded from ascelibrary.org by Kookmin University on 10/14/22. Copyright ASCE. For personal use only; all rights reserved.

considering the compatibility relation between displacements in

共Black et al. 1980兲 共with permission兲
the local and global reference system and noting that the rotation
transformation is independent of displacements. The stiffness ma-
trix of the element in the local reference system is the derivative
of the end forces p̄ with respect to the end displacements ū in this particular, steel strains at the critical sections of the brace. The
system. With the aid of Eq. 共1兲, one obtains latter are very important in low cycle fatigue studies for the pre-
diction of the ultimate fracture of the concentric braces 共Uriz
⳵ p̄ ⳵ ⳵ bu ⳵q ⳵v 2005兲. The issue of mesh refinement is addressed in the paramet-
k̄ = = 共buq兲 = q + bu 共4兲
⳵ ū ⳵ ū ⳵ ū ⳵ v ⳵ ū ric studies of this paper. An inelastic frame element with moderate
deformations overcomes the need for brace subdivision into more
From Eq. 共4兲 it is clear that the element stiffness in the local than two elements. Such an element is not used in the present
reference system consists of two contributions: study, but details of its formulation can be found in de Souza
共a兲 the geometric stiffness matrix k̄g 共2000兲. Note that a middle node is essential in the corotational
formulation of any type of element in order to represent the large
⳵ bu
k̄g = q 共5兲 in- and out-of- plane displacements at the middle of the brace.
⳵ ū The inelastic beam-column element in this study accounts for
that represents the change of the equilibrium transformation ma- the interaction of axial force and bending moment along the brace
trix with respect to the element end displacements; and by integration of the uniaxial, hysteretic steel material model over
共b兲 the basic element stiffness matrix transformed to the local the cross section of the brace 共fiber model兲. The hysteretic model
reference system of Menegotto-Pinto is used for the steel material with extensions
for kinematic and isotropic hardening 共Filippou et al. 1983兲. In a
⳵q ⳵v distributed inelasticity model, the section response is monitored at
k̄m = bu 共6兲
⳵ v ⳵ ū several points along the element axis, but this is not very impor-
tant in the absence of element loads under small deformation
where ⳵q / ⳵v⫽basic element stiffness matrix k. It can be shown theory, because the most critical internal force combinations take
that with a consistent selection of deformation measures place at the end sections. Thus, a beam-column element with
⳵v finite hinge length may be more economical. The integration of
au = bTu = 共7兲 nonlinear material response over the cross section of the brace
⳵ ū
member provides detailed information on inelastic strains, so that
so that the second contribution is symmetric as long as the basic the eventual fracture of the steel brace under low-cycle fatigue
stiffness matrix is. The geometric stiffness contribution is also can be predicted with reasonable confidence 共Uriz 2005兲, but this
symmetric. Expressions for this matrix for the exact theory and will be addressed in a future paper.
different approximations can be found in Filippou and Fenves The inelastic beam-column model in this study is based on the
共2004兲. assumption that plane sections remain plane after deformation
This brief overview demonstrates that if the element formula- and that the section shape is retained. This may not be the case for
tion is made in the basic reference system without rigid body slender cross sections, particularly tubular sections, where the
modes, nonlinear geometry effects can be readily incorporated compatibility of flange deformation provides restraint on the web
during the transformation of the basic response quantities to the deformation. In such a case, the strain in the cross section in the
global reference system of the structural model. This decoupling vicinity of the buckle is much larger than that of a compact
of the element formulation from the geometric transformation of section with an insignificant amount of buckling. Under this cir-
the element response quantities is a significant advantage of the cumstance, modeling of local buckling becomes important in ap-
corotational theory for frame elements and is reflected in the proximating the instant of brace fracture due to low-cycle fatigue.
implementation of geometric transformation classes in OpenSees The correlation studies demonstrate that the hysteretic response of
by R. de Souza 共de Souza 2000兲. the brace is not affected by this effect as long as the section is
The inelastic beam-column element with small deformation compact. In this regard, it should be noted that the seismic pro-
theory in the present study relies on the corotational theory to visions of AISC limit local buckling in concentrically braced lat-
represent the moderate to large deformation effects of inelastic eral load resisting systems in regions of high seismic risk. These
buckling of the concentric brace. To this end, the brace needs to provisions postulate flange/web slenderness ratios of 共b − 2t兲 / t
be subdivided into at least two elements, while a finer subdivision ⬍ 110/ Fy for HSS sections 共AISC 1997兲 and, thus, help extend
is necessary for representing accurately local deformations and, in the fatigue life of steel braces by limiting local strains.

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2008 / 621

J. Struct. Eng., 2008, 134(4): 619-628

Downloaded from ascelibrary.org by Kookmin University on 10/14/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Model geometry and loading for parameter study of 4 in. extra-strong pipe specimen with pin-pin conditions: 共a兲 section discretization;
共b兲 setup; and 共c兲 displacement history

Parametric Study of Brace Buckling elements in the model are indeed small. The larger the number of
elements, the more accurate the representation of the second order
Before embarking on the correlation studies of the proposed effect along the brace. The mesh needs to be more refined for
model with available experimental results, a thorough parametric accurate representation of local response, while a coarse may suf-
study of the typical behavior of a single brace is conducted. This fice for the accurate representation of the overall hysteretic axial
parametric study addresses the effect of modeling parameters on force-displacement relation of the entire brace.
the analytic hysteretic response and, thus, provides guidance for For inducing buckling in an axially loaded brace, it is neces-
the later simulations of actual specimens. A previously tested strut sary to include an imperfection to the geometry of the system in
共Black et al. 1980兲 was used as benchmark for the parametric the form of initial camber, or to the properties of the member in
study. The strut consisted of an extra-strong 4 in. steel pipe with the form of a residual stress distribution over the cross section.
material yield strength of 24 ksi. The length of the brace element Hollow structural members exhibit significant residual stresses as
was 118.4 in., while end support conditions approximated pin a result of the fabrication process 共Sherman 1976兲. These can
conditions in the test, so that the theoretical slenderness ratio of influence the initial buckling load, the cross-sectional deforma-
the brace was 80. This strut geometry was selected on account of tion, and the local buckling. They do not, however, affect in a
the rather compact cross section and the minimal effect of local significant way the hysteretic behavior of a compact section.
and lateral torsional buckling. Fig. 2 shows the measured hyster- Thus, it was decided to neglect this effect in the present study and
etic axial force-displacement relation of the brace. induce buckling through the specification of initial camber.
The analytical model of the brace consists of several inelastic A parabolic initial camber is used throughout this study by
beam-column elements with small deformation theory in the basic specifying the nodes of the structural model accordingly. The re-
system, but large displacement corotational geometry transforma- sults from a finer element discretization show better agreement,
tion of the basic system response to the global reference system of since the elements themselves are straight. The magnitude of the
the structural model. This mesh discretization ensures that the initial camber displacement at midlength of the brace varied be-
deformations of the actual member relative to the chord of the tween 0.01% and 3% of the original brace length. Clearly, the

Fig. 4. Effect of initial deformation on buckling and hysteretic behavior of strut

622 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2008

J. Struct. Eng., 2008, 134(4): 619-628

Downloaded from ascelibrary.org by Kookmin University on 10/14/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Effect of number of elements on hysteretic response of strut for different boundary conditions

larger value is not realistic but was used here as an extreme case. the monotonic response, the subsequent cycles of the hysteretic
The boundary conditions of the structural model, the loading response of the brace in Fig. 4共b兲 are not affected much by the
and its history, and a schematic discretization of the tubular sec- maximum initial camber value. The excellent agreement between
tion are given in Fig. 3. The discretization of the tubular section the numerical results of the hysteretic brace behavior and the
consists of a mesh of 2 ⫻ 30 fibers in the radial and circumferen- corresponding experimental measurements in Fig. 2 is notewor-
tial direction, respectively, for a total of 60 fibers. The material thy. The model is capable of capturing the initial buckling load,
model for structural steel is based on the Menegotto-Pinto model the rapid strength reduction in the postbuckling range, the tensile
with yield strength of 24 ksi, Young’s modulus of 29,000 ksi, and strength with its gradual increase due to strain hardening, and the
strain hardening ratio of 0.3%. The boundary conditions represent residual compressive strength and its evolution during subsequent
pin supports at the ends of the brace. One end was loaded with the cycles of gradually increasing magnitude. It is clear from this
axial force, while the other was restrained from translating in comparison that residual stresses and local buckling do not affect
the axial direction. The imposed displacement history matches the the axial load-displacement response of the brace, confirming the
history from the testing of the same brace by Black et al. 共1980兲. assumptions made at the start of the parametric study. From the
Fig. 4共a兲 shows the effect of initial camber on the monotonic comparison of numerical and experimental results, it can also be
buckling behavior. The brace is subdivided into four elements, so concluded that the smallest value of initial camber of 0.01% of
as to represent more accurately the shape of the initial camber. the brace length gives the best agreement with the buckling load
Clearly, the buckling load is very sensitive to the value of maxi- observed in the test. This is in agreement with observations from
mum initial camber displacement with a significant reduction of other tests 共Gugerli and Goel 1980; Jain and Goel 1979; Khatib
the buckling load in the initial camber range of 0.01 to 1%, which et al. 1988; Lee and Goel 1987兲 and reflects the high standard of
is typical of practical recommendations for compression mem- care of the experimental setup.
bers. A very sharp decrease of the postbuckling strength is ob- Fig. 5 shows the effect of mesh discretization on the hysteretic
served under a very small initial camber value. In contract with axial force-displacement response of the brace for the case of

Fig. 6. Effect of mesh discretization on curvature distribution for seven integration points per element

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2008 / 623

J. Struct. Eng., 2008, 134(4): 619-628

 theory, the largest internal forces and section deformations take
place at the end sections of the element. While the number of
integration points in each element affects the values of section
deformation through the weight of each integration point, this
effect is minor relative to the effect of element size, which is
decisive in the estimation of section deformations in Fig. 6共a兲.
With an inelastic element that accounts for second order deforma-
tions in the basic system 共de Souza 2000兲, the need for mesh
discretization is significantly reduced, because the integration
points along the element can play an important role. Fig. 6共b兲
shows the even greater importance of mesh discretization in the
accurate determination of local deformations for the case of pin-
fixed support conditions. In this case, the location of the critical
section shifts depending on the node location for the particular
Downloaded from ascelibrary.org by Kookmin University on 10/14/22. Copyright ASCE. For personal use only; all rights reserved.

mesh discretization, and differences in the distribution values are

Fig. 7. Effect of number of integration points on hysteretic response observed even between the cases of ten and 30 elements for the
with four elements brace representation. Thus, a finer mesh discretization with more
than ten elements per brace is warranted in this case, or, the
analyst needs to resort to an inelastic element that accounts for
pin-pin and pin-fixed support conditions. In both cases, the same second order deformations in the basic system 共de Souza 2000兲.
initial camber is used with a maximum value of 0.05% of the Fig. 7 shows the effect of the number of integration points in
brace length. The figure compares the numerical results with two, an element on the hysteretic axial load-displacement response of
four, ten, and 30 elements representing the brace. Seven integra- the brace. The numerical results are obtained with a mesh of four
tion points are used in each element. The results lead to the elements for the brace, even though two elements could have
conclusion that the global response of the brace is not sensitive to sufficed according to the earlier parametric study results. The re-
mesh discretization and that, consequently, two elements suffice sults lead to the conclusion that three Gauss integration points are
for the brace representation. It is worth contrasting the hysteretic sufficient for an accurate representation of the hysteretic buckling
behavior of the brace with pin-pin support conditions in Fig. 5共a兲 behavior.
to the behavior of the brace with pin-fixed support conditions
in Fig. 5共b兲. While the behavior under tensile force is essentially
the same, the higher buckling load for the pin-fixed support con-
dition due to the lower slenderness ratio results in a higher re- Correlation Studies with Experimental Results
sidual compressive strength. This, in turn, leads to the higher
energy dissipation capacity of the brace with pin-fixed support The following correlation studies with struts of different cross
conditions. section shapes intend to illustrate the capabilities and limitations
While mesh discretization is not important for the global re- of the proposed model. These studies cover both compact and
sponse of the brace, it plays a significant role in the determination noncompact sections for different tests at different laboratories
of the inelastic deformations at the critical brace sections, and, over the last 25 years. The brace cross-section shapes in the cor-
thus, ultimately in the estimation of the instant of fracture due to relation studies are shown in Fig. 8: Circular and rectangular tu-
low cycle fatigue. This is illustrated in Figs. 6共a and b兲, which bular sections, wide flange shapes, and double angles. The fiber
depict the effect of mesh discretization on the curvature distribu- discretization in the figure is schematic. The actual number of
tion along the brace for pin-pin and pin-fixed support conditions, fibers used in each analysis will be stated with the specimen. All
respectively. Again, two, four, ten, and 30 elements are used for specimens had pin-pin support conditions and were loaded simi-
the brace discretization. Fig. 6共a兲 shows that ten elements give larly to the brace specimen used in the parametric study 共Fig. 3兲.
essentially the same curvature distribution values as 30 elements. In the correlation studies, the numerical model consisted of
Even four elements appear to give distribution values of sufficient two inelastic beam-column elements with small deformation
accuracy. In this case, the critical section lies at brace midspan, theory. Three integration points were used in each element. An
and mesh discretization only serves the purpose of refining the initial camber displacement of 0.05% of the brace length was
estimate of the largest curvature value. It is worth noting that for consistently assumed at midlength of the brace. The hysteretic
the inelastic beam-column element with small deformation material response of structural steel was represented by the model

Fig. 8. Cross section of specimens for correlation studies

624 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2008

J. Struct. Eng., 2008, 134(4): 619-628

Downloaded from ascelibrary.org by Kookmin University on 10/14/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 9. Hysteretic response correlation of W8 ⫻ 20 strut with bucking before yielding in tension

of Menegotto-Pinto with kinematic and isotropic hardening The buckling load of the specimen is overestimated by the analy-
共Filippou et al. 1983兲. The strain hardening ratio was consistently sis, but this result is very sensitive to the assumed initial camber,
assumed equal to 0.3%, and the Young’s modulus was set equal to as was discussed in the parametric studies. The residual compres-
29,000 ksi for all specimens. The yield strength varied from sive strength of the strut in the analysis shows excellent agree-
specimen to specimen in accordance with the values reported in ment with the experiment. The evolution of tensile strength and
the experimental investigations. hysteretic behavior of the strut also shows very good agreement
with the experimental results, except for the hardening behavior
Struts with Compact Sections in tension during the intermediate load cycles. The latter is di-
rectly dependent on the assumed values of kinematic and isotro-
The experimental results of braces with compact cross section
pic strain hardening of the material model. The identification of
stem from an extensive study at UC Berkeley 共Black et al. 1980兲.
the most suitable set of material parameters for each specimen is
Details about setup and conduct of tests, selection and material
beyond the scope of the present study.
properties of the specimens, as well as about instrumentation, and
measurement of response are available in the report. Figs. 9–11 The agreement between analysis and experiment is equally
show the experimental and analytical response of wide flange good for the W6 ⫻ 20 strut with the smaller slenderness ratio of
braces with slenderness ratios of 120, 80, and 40, respectively. All 80 in Fig. 10. In this case, the strut was loaded in tension first,
wide flange sections were represented in the model with five lay- with yield initiation, and then subjected to a high enough com-
ers across the flange thickness and five layers across the depth of pression force to induce buckling. In spite of this difference in
the web, resulting in a total number of 15 layers. loading history relative to the specimen of Fig. 9, the analytical
The W8 ⫻ 20 strut with the high slenderness ratio of 120 was buckling load is again higher than the experimental measurement,
loaded in compression first, with buckling initiation, before load- indicating the importance of the assumed initial camber value,
ing in tension. The comparison of the experimental with the ana- even for this type of load history. The correlation between analy-
lytical response in Figs. 9共a and b兲 shows excellent agreement. sis and experiment confirms the accuracy of the results of the

Fig. 10. Hysteretic response correlation of W6 ⫻ 20 strut with tension yielding before bucking

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2008 / 625

J. Struct. Eng., 2008, 134(4): 619-628

Downloaded from ascelibrary.org by Kookmin University on 10/14/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 11. Hysteretic response correlation of W6 ⫻ 20 strut with small slenderness ratio kl/ r = 40

analytical model regarding residual compressive strength, hyster- between the assumption about the initial camber value in the
etic behavior and tensile strength evolution, as was the case for analysis and the actual test conditions; 共b兲 for the double angle,
the more slender specimen of Fig. 9. the agreement between analysis and experiment regarding the
Similarly good agreement is evident from the comparison of residual compressive strength is less satisfactory; and 共c兲 there
experimental and analytical results for the W6 ⫻ 20 strut with is a sudden reduction in the analytical tensile strength after the
the small slenderness ratio of 40 in Fig. 11. The agreement first postyield excursion, which is not evident in the experimental
of analysis and experiment regarding residual compressive results.
strength, hysteretic behavior, and tensile strength evolution is Fig. 13 shows the comparison of analytical with experimental
very satisfactory. results for a hollow HSS 4 ⫻ 4 ⫻ 1 / 4 strut. In the discretization of
Fig. 12 shows the comparison of the analytical with the ex- this section, six layers are used across the flange thickness and six
perimental results for a L6 ⫻ 3 1 / 2 ⫻ 3 / 8 double angle strut with layers across the depth of the web for a total of 18 layers. The
the intermediate slenderness ratio of 80. In the experiment, agreement between analysis and experiment regarding buckling
stitch plates were used to create a code compliant section, while load, residual compressive strength, tensile strength evolution,
the analytical model assumes perfectly welded angles. The sec- and hysteretic strength of the brace is excellent.
tion is represented in the model with five layers through the flange This series of correlation studies leads to the conclusion that
thickness and five layers across the long legs of the angle. The the proposed analytical model is able to capture with very satis-
agreement of the hysteretic response between analysis and experi- factory accuracy the buckling strength and the hysteretic behavior
ment is similar to what was observed earlier for the wide flange of struts with compact cross sections. The excellent agreement
strut with the same slenderness ratio except for the following between analytical and experimental results confirms the minor
differences: 共a兲 for the double angle strut, the overestimation of role of local buckling in struts with compact cross sections at least
buckling load is relatively small, indicating a better agreement with regard to the overall hysteretic force-displacement relation.

Fig. 12. Hysteretic response correlation of L6 ⫻ 3 1 / 2 ⫻ 3 / 8 hollow strut

626 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2008

J. Struct. Eng., 2008, 134(4): 619-628

Downloaded from ascelibrary.org by Kookmin University on 10/14/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 13. Hysteretic response correlation of HSS 4 ⫻ 4 ⫻ 1 / 4 hollow strut

The correlation studies in this paper do not address the issue of compressive strength upon initiation of buckling. Local buckling
brace fracture as a result of low cycle fatigue, in which case local of the tube walls is the key cause of this behavior. As demon-
buckling plays a more prominent role. This subject is addressed in strated in the analytical results in Fig. 14共b兲, the proposed model
detail elsewhere 共Uriz 2005兲. captures well the buckling strength of the specimen, as well as its
tensile strength. It is, however, incapable of simulating the post-
Struts with Noncompact Sections buckling behavior and the gradual reduction in compressive
strength with increasing number of inelastic excursions in com-
To illustrate the effect of local buckling on the hysteretic axial pression. The assumption of the proposed model that cross sec-
force-displacement relation of noncompact sections, Fig. 14 tions retain their original shape is responsible for the discrepancy
shows the experimental and analytical results of a HSS between analysis and experiment in the postbuckling range of
4 ⫻ 4 ⫻ 1 / 8 hollow strut with noncompact section. The slender- response.
ness ratio of the strut is relatively small with a value of 43. The
slenderness of the tube walls is relatively high with a value of 30.
Local buckling of the tube is rather prominent in this case, as
reported in the experimental studies of Lee and Goel 共1987兲. The Conclusions
report contains details about setup and conduct of tests, selection
and material properties of the specimens, as well as about instru- The paper presents a model for the inelastic buckling behavior of
mentation, and measurement of response, so that these are not steel braces. The model consists of a force-based beam-column
repeated here. Fig. 14共a兲 shows the rapid reduction of the residual element with distributed inelasticity and fiber discretization of the

Fig. 14. Hysteretic response correlation of HSS 4 ⫻ 4 ⫻ 1 / 8 hollow strut

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2008 / 627

J. Struct. Eng., 2008, 134(4): 619-628

 cross section. With this approach the interaction between axial ment inelastic analysis of frames.” Ph.D. thesis, Univ. of California,
force and bending moment is accounted for through integration of Berkeley, Calif.
material response over the cross section. Even though the element Filippou, F. C., and Fenves, G. L. 共2004兲. “Methods of analysis for
only accounts for small deformations in the basic system, large earthquake-resistant structures.” Earthquake engineering: From engi-
displacement geometry is included in the nonlinear transforma- neering seismology to performance-based engineering, Y. Bozorgnia
tion of the force-deformation relation of the basic element follow- and V. V. Bertero, eds., CRC, Boca Raton, Fla.
Filippou, F. C., Popov, E. P., and Bertero, V. V. 共1983兲. “Effects of bond
ing the concept of the corotational formulation.
deterioration on hysteretic behavior of reinforced concrete joints.”
Correlation studies with available experimental results of the
UCB/EERC-83/19, Earthquake Engineering Research Center, Univ. of
hysteretic axial force-displacement response of steel struts dem- California, Berkeley, Berkeley, Calif.
onstrate the ability of the proposed model to represent accurately Gugerli, H., and Goel, S. 共1980兲. “Large scale tests for the hysteresis
the buckling strength, the postbuckling behavior, the tensile behavior of inclined bracing members.” Proc., 7th World Conf. on
strength, and the hysteretic behavior of struts with compact cross Earthquake Engineering, Istanbul, Turkey, 87–94.
section. Hall, J., and Challa, M. V. R. 共1995兲. “Beam-column modeling.” J. Eng.
Through parametric studies of the hysteretic behavior of a tu- Mech., 121共12兲, 1284–1291.
Downloaded from ascelibrary.org by Kookmin University on 10/14/22. Copyright ASCE. For personal use only; all rights reserved.

bular brace, the following modeling recommendations are offered Ikeda, K., and Mahin, S. A. 共1986兲. “Cyclic response of steel braces.”
for the use of this model in professional practice: J. Struct. Eng., 112共2兲, 342–361.
• The brace member should be subdivided into two inelastic Ikeda, K., Mahin, S. A., and Dermitzakis, S. N. 共1984兲. “Phenomenologi-
beam-column elements. cal modeling of steel braces under cyclic loading.” UCB/EERC-84/09,
• An initial camber displacement of 0.05–0.1% of the brace Earthquake Engineering Research Center, Berkeley, Calif.
length should be specified at brace midspan. Jain, A. K., and Goel, S. 共1979兲. “Seismic response of eccentric and
concentric braced frames with different proportions.” UMEE 79R1,
• The inelastic response of each element should be monitored at
College of Engineering, Univ. of Michigan, Ann Arbor, Mich.
three integration points. Jin, J., and El-Tawil, S. 共2003兲. “Inelastic cyclic model for steel braces.”
• Ten–fifteen layers across the depth of the cross section suffice J. Eng. Mech., 129共5兲, 548–557.
for the accurate representation of the spread of yielding and Khatib, F., Mahin, S. A., and Pister, K. S. 共1988兲. “Seismic behavior of
the estimation of inelastic strains at the critical sections of the concentrically braced steel frames.” UCB/EERC-88/01, Earthquake
brace. From these, three–five layers should be used in the Engineering Research Center, Univ. of California, Berkeley, Calif.
flanges and the rest in the web of the cross section. Lee, S., and Goel, S. C. 共1987兲. “Seismic behaviour of hollow and con-
For struts with compact sections, the effect of local buckling crete filled square tubular bracing members.” UMCE87-11, College of
on the overall hysteretic force-displacement behavior of the brace Engineering, Univ. of Michigan, Ann Arbor, Mich.
is not significant. Local buckling is, however, important for the Liu, Z., and Goel, S. 共1988兲. “Cyclic load behavior of concrete-filled
estimation of inelastic strains at critical sections of the brace. tubular braces.” J. Struct. Eng., 114共7兲, 1488–1506.
McKenna, F. 共1997兲. “Object oriented finite element programming frame-
Without the accurate estimation of local strains, it is not possible
works for analysis, algorithms and parallel computing.” Ph.D. thesis,
to predict the brace fracture due to low cycle fatigue. Moreover,
Univ. of California, Berkeley, Calif.
local buckling affects the overall hysteretic behavior of braces Sherman, D. R. 共1976兲. “Application of steel tubing and pipe.” American
with noncompact cross section. For these reasons, it is important Iron and Steel Institute, Washington, D.C.
to extend the proposed model in future studies to account for the Spacone, E., Filippou, F. C., and Taucer, F. F. 共1996兲. “Fiber beam-
effect of local buckling. column model for nonlinear analysis of RC frames. I: Formulation.”
Earthquake Eng. Struct. Dyn., 25共7兲,711–725.
Uriz, P. 共2005兲. “Towards earthquake resistant design of concentrically
References braced steel buildings.” Ph.D. thesis, Univ. of California, Berkeley,
Berkeley, Calif.
AISC. 共1997兲. “Seismic provisions for structural steel buildings.” Ameri- Uriz, P., and Mahin, S. A. 共2004兲. “Physical modeling and seismic vul-
can Institute of Steel Construction, Chicago. nerability assessment of concentrically braced steel frames.” Proc.,
Black, G. R., Wenger, B. A., and Popov, E. P. 共1980兲. “Inelastic buckling Annual Convention, Monterey, Calif.
of steel struts under cyclic load reversals.” UCB/EERC-80/40, Earth- Zayas, V. A., Popov, E. P., and Mahin, S. A. 共1980兲. “Cyclic inelastic
quake Engineering Research Center, Berkeley, Calif. buckling of tubular steel braces.” UCB/EERC-80/16, Earthquake En-
de Souza, R. M. 共2000兲. “Force-based finite element for large displace- gineering Research Center, Berkeley, Calif.

628 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2008

J. Struct. Eng., 2008, 134(4): 619-628